1. Field of the Invention
The present invention relates to transmission of encoded data in a telecommunications system, and, more particularly, to generating signal constellations by a modem for voice or voiceband data transmission.
2. Description of the Related Art
Telecommunication systems commonly employ modulation and encoding of analog signals prior to transmission through a network. Such analog signals are typically voice or voiceband data signals. Voice signals are generated by modulating an electrical signal by the acoustic (voice) signal, while voiceband data signals are generated by modulating an electrical signal such as a carrier with the data. Pulse modulation may then be employed to combine the analog signal with discrete, unit-amplitude pulses before transmission over a telecommunication channel. In pulse amplitude modulation (PAM), the analog signal varies the amplitude of the discrete, unit-amplitude pulses, while, in pulse width modulation (PWM), the analog signal varies the length, in time, of the discrete, unit-amplitude pulses. The original pulse stream is relatively easy for a receiver to detect and regenerate from the signals received from and ideal telecommunication channel. However, since noise and line attenuation/distortion of a real transmission medium (also called the channel response) alters the pulse-modulated signal as it passes through the medium, telecommunication systems typically employ digital transmission techniques. One such digital transmission technique is pulse code modulation (PCM), in which the analog signal is sampled and quantized using discrete digital levels. Typically, 2n discrete levels are employed in telecommunication systems (e.g., using 8 bits, n=8, allowing for 256 discrete levels, with the distance between levels termed quantizing steps).
For a given method of quantizing, each sample of the analog signal is approximated to the nearest discrete level, and the digital value representing the level is transmitted to the receiver. However, since the amplitude of the analog signal and the discrete level of PCM are usually not the same value, the difference between the amplitude of the analog signal and the discrete level of PCM, termed the quantizing error, introduces additional noise into the transmitted signal. This quantizing error introduces noise into the subsequently reconstructed voice or voiceband data signal at the receiver. For PCM using linear quantizing, the increments between the discrete levels are the same (i.e., the quantizing steps are equivalent). However, for linear quantizing, the quantizing noise is not uniform for all analog signal amplitudes because the low amplitude signals experience larger quantizing noise than the high amplitude signals. Consequently, linear quantizing for signals with high dynamic range but with a high percentage of low amplitude signal (such as encoded speech) has a relatively low (poor) signal-to-quantization noise ratio.
Non-linear quantizing with tapered quantizing steps may be employed to compensate for the poor signal-to-quantization noise ratio of linear quantizing. Equivalently, the input signal may be weighted and linear quantizing to achieve the same result. This non-uniform predistortion process, termed companding, compresses larger signal amplitudes, and a receiver then reverses the companding process. Telecommunication systems typically employ a logarithmic companding law. In some countries, such as the United States, PCM line encoding of an analog signal employs a companding function, termed xcexc-law, as given in equation (1):                               e          o                =                                                            log                ⁡                                  (                                      1                    +                                          μ                      ⁢                                              xe2x80x83                                            ⁢                                              e                        i                                                                              )                                                            log                ⁡                                  (                                      1                    +                    μ                                    )                                                      ⁢                          xe2x80x83                        ⁢            0                    ≤                      e            i                    ≤          1                                    (        1        )            
where eo is the output signal value, ei is the normalized input signal value, and xcexc is a constant. Other countries, such as Europe, employ a different companding function, termed A-law, and are given in equation (2):                               e          o                =                  {                                                                                          Ae                    i                                                        1                    +                                          log                      ⁡                                              (                        A                        )                                                                                                                                                              if                    ⁢                                          xe2x80x83                                        ⁢                    0                                    ≤                                      e                    i                                    ≤                                      1                    A                                                                                                                                            1                    +                                          log                      ⁡                                              (                                                  Ae                          i                                                )                                                                                                  1                    +                                          log                      ⁡                                              (                        A                        )                                                                                                                                                              if                    ⁢                                          xe2x80x83                                        ⁢                                          1                      A                                                        ≤                                      e                    i                                    ≤                  1                                                                                        (        2        )            
where eo is the output signal value, ei is the normalized input signal value, and A is a constant greater than 1. Since voice and voiceband data signals are often transmitted between different systems using either xcexc-law or A-law, telecommunication networks provide for reformatting (encoding conversion) between the two companding functions.
Encoding conversion may be between networks employing A-law encoding and networks employing xcexc-law encoding. Such encoding conversion may be implemented within a network as a simple mapping between A-law and xcexc-law levels (i.e., a mapping between A-law and xcexc-law encoded sample values). As would be apparent to one skilled in the art, such mapping may add signal distortion from quantization error. For example, during an initial encoding, samples of an analog signal may be mapped to corresponding xcexc-law levels, which are subsequently mapped to A-law levels during encoding conversion. Since A-law encoding and xcexc-law encoding are non-linear companding methods, two xcexc-law levels may map to the same A-law level. Consequently, quantization error may be added to the original signal reconstructed from the sequence of A-law levels when certain xcexc-law quantizing level information is lost.
For PCM systems in some countries, such as the United States, voice or voiceband data channels are subject to PCM encoding and grouped by time multiplexing into twenty-four 8-bit channels (192 bits). One framing bit is appended to this group of 192 bits to form a T1 format of 193 bits/frame. The pattern of framing bits received over several frames (e.g., twelve T1 frames) may be employed for T1 line framing and timing synchronization, as well as for line bit error rate (BER) calculation. The voice and voiceband data signals are typically sampled at 8 kHz, so each T1 frame is transmitted at 1.544 Mb/sec. Similar formats exist in other countries, such as the E1 frame comprising thirty 8-bit channels plus framing bits transmitted at 2.048 Mb/sec. Signaling for set-up/tear-down of connections, or other slow-speed network data channels, may be superimposed on the T1 frame (termed herein as xe2x80x9csuperimposed information channelsxe2x80x9d). For example, for robbed-bit signaling (RBS), a low-rate signaling channel may be formed by replacing the least significant bit (LSB) of each 8-bit channel with a signaling bit of the low-rate signaling channel for every sixth T1 frame transmitted. For RBS using two signaling bits, the period of the RBS information channel is twelve T1 frames. For RBS using four signaling bits, the period of the RBS information channel is twenty-four T1 frames (the pattern of T1 framing bits over twenty-four frames also defines a T1 superframe).
The present invention relates to detection of transmission line characteristics of a telecommunication channel, such as the companding law used for line encoding, conversion between line encodings having different companding laws, superimposed information channels, or line attenuation, between a pair of transceivers. Such detection may be accomplished through a set of test levels that are transmitted from one transceiver to the other transceiver of the pair. The transceiver receiving the set of test levels transmitted through the telecommunication channel compares normalized received test levels with expected, normalized, ideal values to detect one or more of the transmission line characteristics. Analog signal levels used for the test levels are determined based on the detected line encoding companding laws and the particular transmission line characteristic to be detected. The transceiver receiving the set of test levels and detecting the line encoding and other transmission line characteristics may then correct for distortion of signals caused by the detected transmission line characteristics. Correcting for distortion allows for higher received signal to noise ratio, lower bit error rate, and higher data rate transmission between the pair of transceivers. In accordance with embodiments of the present invention, a modem adjusts the signal representing a constellation to compensate for distortion. A source modem transmits the signal with a constellation based on one line encoding algorithm and a destination modem transmits to the source modem with a constellation based on an other line encoding algorithm. A superimposed information channel in the signal is detected to identify channel compensation values; line attenuation of the signal is detected to identify gain compensation values; and the constellation is adjusted at the destination modem in accordance with the encoding conversion, channel, and gain compensation values.